Intel Achieves Milestone: World’s Largest Neuromorphic Computing System with a Billion Neurons

Intel has made a significant breakthrough in the field of neuromorphic computing by unveiling the world's largest system, known as the Hala Point. This cutting-edge system boasts an impressive 1.15 billion neurons, powered by Intel's Loihi 2 processor. Initially deployed at Sandia National Laboratories, the Hala Point system represents a major leap forward in the realm of artificial intelligence.

Featuring 1,152 Loihi 2 processors produced on Intel 4 process node, the Hala Point system is housed in a six-rack-unit data center chassis. With a capacity of up to 1.15 billion neurons and 128 billion synapses spread across 140,544 neuromorphic processing cores, the system consumes a maximum of 2,600 watts of power. Additionally, it incorporates over 2,300 embedded x86 processors for ancillary computations, making it a powerhouse of computational capabilities.

One of the most impressive aspects of the Hala Point system is its ability to support up to 20 quadrillion operations per second, achieving an efficiency exceeding 15 trillion 8-bit operations per second per watt (TOPS/W) when running conventional deep neural networks. This remarkable performance opens up new possibilities for research into brain-inspired artificial intelligence and addresses the pressing issue of power efficiency in modern AI data center systems.

Compared to its predecessor, the Pohoiki Springs system, the Hala Point system offers a tenfold increase in neuron capacity and up to 12 times higher performance. It is capable of running mainstream AI workloads with ease, thanks to its massively parallelized fabric that enables impressive memory bandwidth, inter-core communication bandwidth, and inter-chip communication bandwidth.

When applied to bio-inspired spiking neural network models, the Hala Point system can operate at speeds up to 20 times faster than the human brain at full capacity and up to 200 times faster at lower capacities. While not designed for neuroscience modeling, the system's neuron capacity is comparable to that of an owl brain or the cortex of a capuchin monkey, showcasing its immense computational power.